This lack of efficient therapies for tinnitus partly originates from an incomplete understanding of the pathophysiology of tinnitus, which is only starting to be unveiled. Whereas an initial event in the inner ear or in peripheral auditory pathways may initiate tinnitus, current hypotheses suggest that tinnitus could be maintained by the development of an abnormal plasticity triggered by deafferentation in a wide network of cortical areas and possibly subcortical structures [6, 23, 32]. Studies in animal models and patients have demonstrated both loss of hair cells in the inner hear and abnormal plastic alterations of tonotopic maps in the auditory cortex [25, 31, 32], with the now classical “edge effect” which could be one of the mechanisms underlying tinnitus. A zone of the auditory cortex that used to represent a frequency band lost by deafferentation could be “invaded” by the adjacent frequency bands, leading to an enduring auditory phantom perception, i.e., to tinnitus [25, 32]. When comparing tinnitus patients with healthy volunteers, functional brain imaging studies using positron emission tomography (PET) or functional magnetic resonance imaging (fMRI) have consistently demonstrated abnormalities of the resting metabolism or activation in several brain structures, including the primary and secondary auditory cortices, limbic and emotional or attentional areas such as the dorsolateral prefrontal cortex, and subcortical structures [23–25]. Whereas the abnormal activity disclosed in the auditory-related areas (primary and secondary auditory cortices) could underlie the phantom auditory perception itself, the abnormal activity observed in the areas supporting cognitive, attentional, and limbic processes could be involved in the unpleasant and distressing aspect of tinnitus [31, 32]. Evoked auditory potentials or magnetoencephalography also point toward the persistence of an abnormal electric activity in a distributed network in tinnitus patients [34, 39].

The discovery of a disordered excitability in the auditory pathways and maladaptive plastic changes in auditory and limbic cortical areas led to the exciting hypothesis that it should be possible to treat tinnitus by modulating these abnormalities through brain stimulations. Indeed, several studies demonstrated that a single session of repetitive transcranial magnetic stimulation (rTMS) induces a transient relief of tinnitus [9, 29]. However, the therapeutic effect of rTMS waned as time elapsed. Therefore, aiming to extend this transient beneficial effect, several teams successfully applied repeated sessions of rTMS and induced a prolonged relief in tinnitus patients [15, 17, 30, 33]. Direct electric brain stimulation with epidural electrodes implanted over the auditory cortical areas has also been applied successfully in tinnitus patients; however, this approach requires invasive neurosurgical procedures [5].

Ten years ago, another method of noninvasive transcranial brain stimulation re-emerged after a long eclipse period: transcranial direct current stimulation (tDCS) [27]. Transcranial direct current stimulation has been extensively used to explore the neurophysiological mechanisms that govern human brain plasticity, as well as for modulating brain excitability. When applied over the primary motor cortex (M1), anodal (cathodal) tDCS increases (decreases) M1 excitability beyond the period of stimulation [27]. Transcranial direct current stimulation is considered as a promising therapeutic tool for several neurological and psychiatric pathologies which are driven by or result from abnormal brain excitability such as stroke, Parkinson’s disease, chronic pain, and depression [10].

The study assumed a double-blind, placebo-controlled, cross-over design. After a baseline evaluation with questionnaires (see below) at the otorhinolaryngology outpatient consultation, each patient underwent three sessions of tDCS (anodal, cathodal, and sham) in a balanced order, at 2 week intervals. Two weeks after the last tDCS session, the patients were evaluated again during a follow-up visit. A total of 21 tinnitus patients (5 females) signed informed consent and were enrolled after a structured interview as suggested by the Tinnitus Research Initiative (TRI) [20].

Transcranial direct current stimulation was delivered by an Eldith DC-Stimulator^® (NeuroConn, Ilmenau, Germany) through electrodes embedded in sponges soaked with NaCl 0.9%. The target cortical area was the left temporoparietal area (LTA), defined as being halfway between C3 and T5 measured with the 10-20 international EEG system [9, 29]. A 35 cm² electrode was positioned over the LTA and a 50 cm² electrode was positioned on the right scalp between T4 and F8. In order to guarantee the double-blind aspect of the experiment, an experimenter initially introduced a real or sham code for each session. The order of the codes was balanced across patients and sessions. A second (blinded) experimenter collected the behavioral data. During real (anodal/cathodal) stimulation, the Eldith DC-Stimulator delivered 20 minutes of DC stimulation at 1 mA (1 mA plateau, fade in/out 8 s). For sham tDCS, the polarity of the LTA electrode was anodal in 50% of the patients; after a short up-ramp, brief current pulses (110 μA over 15 ms, peak current 3 ms) were delivered every 550 ms, eliciting a tingling sensation similar to that felt during real stimulation.

The primary outcome measure was a change in tinnitus intensity or discomfort assessed with a Visual Analogic Scale (VAS) change-scale immediately after tDCS and 1 hour later. The VAS change-scale ranged from “full relief” (+4) to “very strong deterioration” (−4), with 0 being “unchanged”. Secondary outcome measures included potential long-term effects over the 2 weeks following each tDCS session on a French translation of the Tinnitus Questionnaire (TQ) [11, 22] and the Beck Depression Inventory (BDI) [2], as well as free reports from patients. The TQ, BDI, and free reports were collected at inclusion and follow-up consultations, and before each tDCS session. The tinnitus patients were dichotomized into a “low, compensated” (TQ < 46) or “high, decompensated” (TQ ≥ 47) distress category, according to the burden of psychological and depressive features [38].

Transcranial direct current stimulation was well tolerated. One of the patients (#19) reported a transient instability after sham tDCS which resolved within 1 hour. Some patients mentioned transient or longer-lasting tinnitus worsening, mostly under cathodal tDCS (see below).

A global analysis demonstrated a significant effect (p = 0.013) for tinnitus intensity immediately after tDCS. When compared to sham tDCS, the VAS change-scale for tinnitus intensity significantly decreased immediately after anodal tDCS (p = 0.020). There was no statistically significant difference between sham and cathodal (p = 0.414) nor between anodal and cathodal tDCS (p = 0.132). One hour after tDCS, the VAS change-scale for intensity was not different anymore between anodal, cathodal, and sham tDCS (p = 0.543). Whereas, the temporal evolution of the VAS change-scales for discomfort were similar to those for tinnitus intensity (Fig. 1), there was no significant difference immediately after anodal, cathodal, or sham tDCS (p = 0.108), nor 1 hour later (p = 0.761).

While 7 patients were positive responders to anodal tDCS, 4 other patients reported a slight tinnitus worsening after anodal tDCS. Six patients were positive responders to cathodal tDCS; 5 patients were also positive responders to anodal tDCS. Five others patients reported slight tinnitus worsening after cathodal tDCS, 1 patient responded positively to anodal tDCS, and 2 patients experienced worsening after anodal tDCS.

As shown in Fig. 1, the temporal evolution of the variances of the tinnitus intensity and discomfort VAS change-scales were strikingly different under real (anodal and cathodal) and sham tDCS. Indeed, the variance remained virtually unchanged after sham tDCS but increased dramatically after anodal and cathodal tDCS. When compared to sham tDCS, the variances of the intensity VAS change-scales increased immediately after anodal and cathodal tDCS (p = 0.011 and p < 0.001, respectively) and kept increasing 1 hour later (p = 0.004 and p = 0.001). There was no statistically significant difference in the variances between anodal and cathodal tDCS (p > 0.283). For the discomfort VAS change-scale, the variances were not significantly different immediately after tDCS (p > 0.108). However, 1 hour later, the variance significantly increased after anodal (p = 0.004) and cathodal (p = 0.002) tDCS when compared to sham tDCS; again, there was no statistically significant difference between anodal and cathodal tDCS (p = 0.9).

Unexpectedly, several patients spontaneously reported either lasting changes of tinnitus characteristics (loudness, pitch, fluctuations, etc.) or improvement/worsening that persisted several days (see Table 2). After anodal tDCS, 6 patients reported prolonged positive effects, and 3 patients mentioned lasting negative effects. After cathodal tDCS, 2 patients reported a lasting negative effect and 3 patients a positive effect. After sham tDCS, 3 patients reported a protracted negative effect. Overall, patients reported long-lasting benefits more frequently after anodal than sham tDCS (p = 0.031), whereas there was no significant difference between sham and cathodal tDCS (p = 0.25), nor between anodal and cathodal tDCS (p = 0.375).

Two key differences with previous tDCS studies [9, 37] were the experimental design and the electrode montage. First, the current experiment assumed a double-blind, placebo-controlled, and cross-over design with a long washout period (2 weeks). This temporal aspect is especially important since the recent history of cortical excitability can shape (and even reverse) the direction of excitability changes induced by tDCS or rTMS, this phenomenon is known as homeostatic plasticity or metaplasticity [40]. The second key difference with previous studies, in which the two electrodes were of the same size, is that we positioned a medium (35 cm²) electrode over the left LTA and a larger (50 cm²) electrode over the right laterofrontal scalp. Recent tDCS experiments have demonstrated that the excitability of a cortical area under a large electrode remains unchanged, ensuring that this large electrode could be considered as a real “neutral” reference electrode [26]. Using a smaller electrode targeting the cortical area of interest and a large reference electrode helps to disambiguate the interpretation of previous tDCS studies in which a combined effect from both “active” electrodes with opposite polarities could not be ruled out. Another way to avoid this ambiguity is to use an extracephalic reference electrode, which seems to be a safe alternative [35]. In the current experiment, the effects observed on tinnitus intensity may be chiefly driven by the DC stimulation of the left LTA; however, a significant contribution of the right frontal cognitive and limbic areas involved the emotional aspects of tinnitus [31] can not be ruled out.

Both our responder rate (35% of the patients were “positive responders” to anodal tDCS) and effect size were comparable to those reported in previous rTMS [19, 28] and tDCS studies [9, 37]. Whereas relatively low, these responder rate and effect size still compare favorably with the lower rate of success of other therapeutic approaches. Moreover, it should be acknowledged that the therapeutic trials designed to modulate tinnitus with noninvasive transcranial brain stimulations are still in their infancy and that the optimal parameters (target area, intensity, duration, etc.) remain to be settled before launching large randomized control trials. Despite the current lack of such trial, some patient’s characteristics seem to predict a positive response to rTMS, such as tinnitus duration and normal hearing [18], whereas age, gender, and tinnitus duration, type and laterality are not predictive for tDCS [37].

Whereas a significant beneficial effect was observed exclusively for anodal tDCS, it is, however, striking to compare the stability of the variance under sham tDCS with the dynamical evolution of the variances after both anodal and cathodal tDCS (Fig. 1). Two hypotheses could explain this observation. The first one is that patients became aware of the real/sham nature of the stimulations. In order to avoid drawing their attention on this aspect, we kept talking with them during stimulation. We did not ask specifically what their guess was; therefore, this possibility cannot be formally excluded. However, this is unlikely for several reasons. First, the stimulation parameters used are considered as “classical” in tDCS experiments and have been proven safe for patient’s blinding, although a significant proportion of healthy volunteers may formulate a correct guess at 1 mA [1]. Second, during sham tDCS, the Eldith DC-Stimulator^® delivered brief and ineffective current pulses inducing a similar itching sensation that improves the blinding procedure. Third, the order of the stimulations was balanced across patients and sessions, which decreases the likelihood of unblinding all the patients. Fourth, the behavioral data were collected in a double-blind fashion. Finally, after sham tDCS, 4 of 20 patients reported short-lasting effects and 3 longer-lasting effects, suggesting they did not guess the sham nature of that stimulation. Thus, it is unlikely that the differential temporal evolution of the variance between sham and real tDCS could be explained by an unblinding of the patients.

The second hypothesis to explain this increasing variance over time solely after real but not sham tDCS is that both anodal and cathodal tDCS interacted with the left LTA activity and triggered changes in tinnitus perception/discomfort which built up over at least 1 hour, but with divergent effects (favorable or unfavorable) in different patients. At the group level, in addition to an increase of variance over time suggesting an ongoing dynamical phenomenon, anodal tDCS resulted in a net favorable effect, whereas cathodal tDCS did not. Besides the “canonical” effects on M1 excitability in humans [27], anodal and cathodal tDCS may both have interfered with an ongoing state of abnormal excitability such as observed in tinnitus patients [31, 32]. This may be in line with the recent observation of large-scale network abnormalities as revealed by functional connectivity in tinnitus patients [36]. Given the fact that several patients also reported additional changes in their tinnitus features (pitch, loudness, laterality) after real tDCS, one could hypothesize that anodal and cathodal tDCS perturbed a maladaptive equilibrium state and opened the door for plastic changes to develop, some patients taking advantage and experiencing an improvement, while others reported a worsening or useless changes in tinnitus features. The reason why tDCS triggers beneficial changes in some patients and detrimental in others remains an open question; reminiscent of the issue why some patients respond to rTMS or tDCS and others do not. This could be related to the clinical history and tinnitus features [18, 37], to the particular excitability of network reorganization (more or less “limbic” or “auditory”) [31, 32], the recent history of the LTA or connected areas excitability changes (homeostatic plasticity) [40], or to their genetic background such as the polymorphism of the brain-derived neurotrophic factor [4].

If real tDCS triggered spontaneous plastic changes by disrupting an ongoing state of abnormal excitability, this could also potentially explain the unexpected long-term effects on tinnitus spontaneously reported by our patients. Recent studies with repeated rTMS sessions have demonstrated a remarkably high success rate with long-term benefits [16]. It is worth noting that long-lasting beneficial effects on tinnitus have been induced by repeated sessions of either low-frequency (“inhibitory”) rTMS or high-frequency (“excitatory”) rTMS in a study with a parallel group design [13–15], suggesting that both low- and high-frequency rTMS could trigger plastic changes in tinnitus patients, just as we observed with a single session of anodal or cathodal tDCS.